BlackoutIntelligence Legal · Corporate · Government Channels on standby Proprietary & Confidential  /  Public record sources only
Blackout Intelligence · Declassified Document Search

THE
BLACKFILES

Operate with clarity. Strike with certainty.

Enter and the console begins acquiring immediately — thousands of declassified records streaming in from the public archives that hold them, newest first. Every record carries a link back to its originating archive. CIA Reading Room documents are searched by the words printed inside them, not only their titles. Nothing is reconstructed, nothing is inferred, and no evidence grade is ever assigned by the machine.

Request a confidential briefing
Acquisition begins the moment you enter and keeps running · nothing is pre-loaded
Testing live channels…
or go straight to a subject
Try
TraceableLive link back to the originating archive on every record.
AttributableAdmiralty grading is entered by a researcher, never by the system.
DefensibleUnreachable archives are reported as failed, never back-filled.
Search channels —
0Records held
0Agencies
0Countries
Live search across public declassification archives Sourced exclusively from public records No fabricated evidence © Blackout Intelligence & Surveillance
THE BLACKFILES / Blackout Intelligence
⌘K
0 held
Idle
RS
R. Sinclair
Researcher
Blackout Intelligence · The Blackfiles

Command

The state of the corpus and your investigation at a glance. Every figure is computed from records actually held — nothing here is estimated or synthetic.
The document index Held under public disclosure · processed for research access
Holdings

Held records

Everything currently in this session's index — offline reserve, live search results you chose to keep, and anything imported through the ingestion connectors.
Records held
0
indexed this session
Agencies covered
0
distinct originators
Graded A or B
0
Admiralty reliability
Pending review
0
awaiting a researcher's grade
Acquisition idle
0 acquired  /  0 examined
Acquisition runs continuously in the background for as long as this console is open. It works through a queue of seventy subjects one at a time, paging each to exhaustion before moving on, so holdings keep climbing while you search, read and grade. Records land as Pending review with no evidence grade. A channel that refuses is reported as refused and never substituted with invented data.
Now working — 0 of 0 topics complete
Throughput0records a minute
Time to finish—rough, from the average so far
Subject queue Click a subject to search it now · the queue works top to bottom
Standby.
Idle
Document index
Title▼ Agency▼ Type▼ Released▼ Grade▼
Every row here came from a named public archive and keeps its original source link. The offline reserve is 2,174 records transcribed from the Record Identification Form printed on each PDF in the NARA 2025 JFK release — where the form carried no readable title, the record says so rather than inventing one. Records arrive as Pending review and stay there until a researcher assigns an Admiralty grade. The system never assigns one for you.
Working sets

Case files

Group records into a named case file, then export the whole set as evidence-ready CSV or a printable dossier.
Priority

Flagged records

Everything you've flagged during this session, newest flag first.
Chronology

Release timeline

Holdings by year. Switch between the year the document itself carries and the year it reached the public — they are rarely the same, and the gap between them is often the point.
Originators

Agencies

Click any originator to filter the held index down to its records.
Jurisdictions

Countries

Where each holding was released, by the archive that published it.
Link analysis

Entities & links

The people, codenames and organisations named across the holdings, and how they co-occur. This is read from what the records themselves say — an entity exists here only because a record names it, and a link means two entities appear in the same record, not that any real-world connection is claimed. Click any entity for its connection graph.
Standing watch

Watchlist

Queries the acquisition engine checks every incoming record against. When a new record matches — a name, an operation, a file series — it is counted here as it arrives, so a subject you care about surfaces the moment it enters the holdings.
Filing context

Archive boxes

Records released under the JFK Records Act carry a group–box–item number. Everything sharing a box was filed together by the originating agency, which is a relationship no keyword search will surface. Open a box to see what travelled with what.
Subjects

Name index

Personal names detected in the titles of held records, in surname-forename form. This is pattern matching on titles, not a curated who's-who: a name only appears here if a record title carries it, and records whose title was never stated on the form cannot contribute. Click a name to sweep every archive for it.
Known operations and inquiries
Provenance

Archive directory

Every archive this console can reach, what it holds, and how it responds to a browser. Archives marked hand-off refuse cross-origin requests by policy — the console carries your query to their own search page instead of pretending to have queried them.
Reach

Corpus & harvest

How much declassified paper is actually reachable from this console, and how to pull it down at volume. Every total below is reported by the archive itself — none of it is estimated, and an archive that will not answer is shown as unreported rather than counted as nothing.
Bulk harvest
archive.org scrape channel · no key · cursor-paginated

The bulk channel pages continuously through a result set rather than returning a single page, so holdings can go from thousands into the hundreds of thousands. Records are written to the browser vault as they arrive and survive a reload — run this a few times on different queries and the corpus accumulates. Everything lands as Pending review; grading remains a human judgement at any volume.

Standby 0 pulled · 0 new · 0 duplicate · archive reports — matching
The vault
browser-resident, persists between sessions

A single HTML file cannot carry millions of records — a million would be a 168 MB download before a pixel rendered, and CIA CREST alone would run past two gigabytes. The vault sidesteps that: records are stored by the browser against your disk quota rather than baked into the file, so the corpus grows across sessions instead of resetting. Records beyond the in-session ceiling stay in the vault and remain exportable.

Session

Search history

Every sweep you've run, with what each archive returned. Re-run any line to repeat it exactly.
Acquisition

Ingestion

Bulk-load records into the index from a catalogue API, or paste structured data from an archive that can't be queried directly.
Internet Archive — bulk declassified collections
Queries archive.org's public search index straight from your browser: open, no key, no registration. Real items only — title, best-available year, and a link back to the archive.org item page. Use this connector when you want volume in a few clicks.
archive.org/advancedsearch.php · no key · up to 500 per batch
Preset:
A live query against a real public index — counts and titles depend on what archive.org has catalogued. Its metadata carries no intelligence grade, so every record lands as Pending review whatever the batch size.
DocumentCloud — FOIA releases published by newsrooms
The public side of DocumentCloud, where the National Security Archive, MuckRock and investigative newsrooms post the documents they've prised loose. Often holds releases that never appear in an agency's own reading room.
api.www.documentcloud.org · no key for public documents
National Archives Catalog (US) — free key required
NARA's own catalogue holds the largest single body of digitised US declassified records. Unlike Internet Archive it needs a personal key: request one free from Catalog_API@nara.gov and paste it below. Field mapping is best-effort — read the log after your first fetch.
catalog.archives.gov/api/v2 · free key · 10,000 queries/month
The key is used only for this in-browser request. It is not transmitted anywhere else and is not written to storage. Custom headers trigger a CORS preflight that some networks block; if it fails, Internet Archive above is the reliable fallback for volume.
CourtListener — published court opinions
Imports real published opinions with their case name, court, filing date and a link to the official document. Useful for surveillance and FISA litigation. Most endpoints now want a free API token — paste one if your request is refused.
courtlistener.com/api/rest/v4 · token optional
Required: title, agency, type, dateReleased. sourceUrl strongly recommended.
First row must be a header matching the JSON field names.
Method

How this works

The sweep

A query is sent to every reachable archive at the same time. Each one answers independently: a count, an empty result, or a failure. Failures are shown as failures. Nothing is filled in from another source to make the board look complete.

Why some archives only get a hand-off link

CIA CREST, the FBI Vault, GCHQ and several national archives set a cross-origin policy that stops a browser reading their results directly. Rather than claim a result it can't produce, the console builds the same query as a link into that archive's own search page and opens it in a new tab.

Admiralty grading

Records are graded on the NATO Admiralty scale, which separates two judgements that get confused constantly:

A–F
Reliability of the source — how much the originating archive or body has earned trust.
1–6
Credibility of the information — whether the content is confirmed by other material.

A record can be A6: an impeccable source carrying a claim nothing else supports. The system will not assign either value. It records who assigned it and when.

What is stored

Everything lives in your browser for this session — flags, notes, grades, case files. API keys are used for the single request you make and are never written to storage. Export before you close the tab if you want to keep a working set.

↑↓ Move↵ SelectEsc Close